584results about "Web data querying" patented technology

The invention relates to artificial intelligence security, in particular to a large model security alignment method based on intention reasoning and RAG guidance, which comprises the following steps: 1, performing structured enhancement on original input before large model reasoning, including security criterion introduction and prison break strategy matching; 2, on the basis of input enhancement, in combination with thinking chain reasoning and a multi-path sampling mechanism, fine-grained classification is carried out on user intentions, and illusion and misjudgment in the reasoning process are reduced; and 3, after the large model generates a response, performing multi-dimensional security assessment on the output by means of an external assessment tool, and guiding the large model to perform iterative optimization based on feedback to form a'generation-assessment-correction 'closed-loop mechanism, the method effectively resists prison break attacks, corrects the inherent defects of the large model, such as illusion, prejudice and the like, and improves the security of the large model. And by displaying the reasoning path and the improvement process, the interpretability of large model decision is enhanced, the general compliance constraint is broken through, and the precise compliance guarantee of high-sensitivity scenes such as medical treatment and finance is realized.

Systems and methods for retrieving information from a corpus of documents that is relevant to a query. The method comprises: generating a first plurality of chunks by subdividing each document in the corpus of documents into one or more chunks of a first size; generating a second plurality of chunks by subdividing each document in the corpus of documents into one or more chunks of a second, larger, size; using an information retrieval system to identify, from the second plurality of chunks, a set of chunks of the second size that are relevant to a query; and using the information retrieval system to identify, from a subset of chunks of the first plurality of chunks, a set of chunks of the first size that are relevant to the query. The subset is based on the set of chunks of the second size that are relevant to the query.

The invention provides an RAG adaptive evidence refining method based on submodule optimization, and belongs to the technical field of artificial intelligence. Collecting data to construct a knowledge base, and generating a candidate document set according to the knowledge base inquired and retrieved by a user; designing an instruction and calling a DeepSeek model to evaluate the answer contribution degree of the candidate document, calculating the relevancy between the query and the candidate document and the information coverage degree score of the optimal document to the candidate document, and constructing a sub-module objective function; calculating a marginal gain threshold value, iteratively calculating optimal documents with marginal gains greater than the threshold value, and generating an optimal document set; and splitting the optimal document set according to the paragraph level, generating a candidate paragraph set, calculating the redundancy score of the optimal paragraph set, constructing a sub-module objective function added with redundancy penalty by multiplexing the method, and performing iterative calculation to generate a final optimal evidence paragraph subset. The method effectively solves the problems of over-high redundancy, over-low information coverage and the like caused by a traditional processing method, so that the final model reasoning performance is improved.

Various embodiments improve search technologies and computer information retrieval by executing a query via ranking a set of search result candidates higher than another set search result candidates based at least in part on the query and determining that a first set of search result candidates are indicative of a sub-accessory to an accessory or an accessory itself.

The invention relates to the technical field of micromolecule mineral water monitoring, and discloses a micromolecule mineral water quality monitoring system based on an intelligent sensor. The system comprises a multi-channel spectrum sensor array, an ion selective field effect transistor group, a microfluidic preprocessing module, a dynamic compensation database, a self-adaptive analysis engine, a cloud verification interface and a feedback execution mechanism. The micro-fluidic pretreatment module is used for eliminating bubbles of an original water sample, screening particles and outputting stable liquid flow; the spectrum sensor array captures full-band absorption spectrum data to generate a characteristic matrix, and the transistor group detects concentration fluctuation of calcium ions, magnesium ions and silicate; the adaptive analysis engine performs multi-modal fusion and abnormal marking on the time sequence alignment data stream; the cloud interface performs cross validation on the local result and the regional water quality center reference library to generate a credibility report; and the feedback execution mechanism adjusts the microfluidic flow velocity and the sensor integration time according to a report to realize efficient water quality monitoring.

The invention relates to the field of internet search service, and discloses a big data-based investment attraction clue analysis system, which comprises a query analysis unit, a query analysis unit, a query analysis unit, a query analysis unit and a query analysis unit, wherein the query analysis unit calls an event sequence template for defining a meta event type and a standard relative time window constraint interval; the distributed retrieval and aggregation unit is used for retrieving and aggregating meta event types to entity objects; the sequential logic verification unit is used for counting the global document frequency of the meta event, reversely mapping the global document frequency to generate a zoom coefficient, dynamically correcting a time window, and judging that only an entity object in a dynamic verification interval after correction is passed; the object sorting unit generates sorting scores according to template matching integrity, and by constructing a density and window reverse coupling mechanism, the sequential logic boundary is adaptively compressed in a high background noise environment, and the retrieval signal-to-noise ratio and the logic confidence coefficient are maintained.

Systems and methods for generating and providing outputs in a multi-device system can include leveraging environment-based prompt generation and generative model response generation to provide dynamic response generation and display. The systems and methods can obtain input data associated with one or more computing devices within an environment, can obtain environment data descriptive of the plurality of computing devices within the environment, and can generate a prompt based on the input data and environment data. The prompt can be processed with a generative model to generate a model-generated output. The model-generated output can then be transmitted to a particular computing device of the plurality of computing devices.

Methods, systems, and apparatus, including computer programs encoded on computer storage media for performance of actions on a computing device using a personal artificial intelligence agent. The system receives, via a user interface of a client device, an input related to the execution of a computer-related task, the input comprising a text or an audio signal converted to text. The system determines one or more services, databases or APIs to retrieve data or information related to the received input and asynchronously obtains the data or information. The system determines one or more actions to be performed responsive to received input. The system provides a generated prompt as input to one or more machine learning models comprising a large language model, the received input and / or the determined or more actions. The system performs the determined one or more actions by the client device.

The invention relates to a page display method and device, equipment and a medium in the technical field of computers, and the method comprises the steps: responding to a page display request, and determining a target authorization party type and a page resource path corresponding to the request; executing a corresponding preset authorization process according to the target authorization party type, and obtaining corresponding local user information including basic information and authority information of a user; reading corresponding page resource data according to the page resource path, and injecting local user information into the page resource data; and carrying out a rendering process of matching local user information on the injected page resource data, and returning an obtained page file to the client which sends the page display request. According to the method and the device, efficient page display service with better visual impression is provided.

The present application provides a knowledge base construction method, a network element, and a storage medium. The method comprises: when multi-modal data is not a vector, converting the multi-modal data into a vector on the basis of an embedding model; and storing the vector corresponding to the multi-modal data into a knowledge base of a second network element.

The invention discloses an information fusion and reasoning method and system based on multi-agent collaborative networking search, and relates to the technical field of network information, and the method comprises the steps: a deep research strategy scheduling agent deconstructs an unstructured request through a task formalization mechanism, generates a high-dimensional instruction code, and dynamically optimizes an execution path. And the network intelligence index generation agent extracts and screens the trusted URL by using a rule engine and a semantic evaluator based on the code, and generates a structured link instruction stream. The structured analysis intelligent agent implements multi-level semantic distillation and conversion on an original document to obtain key value pair representations, and the key value pair representations are packaged into machine readable data. And generating a comprehensive intelligence abstract with consistent logic through the information fusion analysis agent. And the natural language generation engine converts the information into a professional report, and returns the professional report to the scheduling module through a feedback interface to realize cognitive closed-loop iterative optimization. The technical problem that in the prior art, it is difficult for geologists to obtain accurate, comprehensive and credible professional conclusions is solved.

A system and method includes: providing a system comprising one or more servers, a user interface communicably coupled to the one or more servers, one or more databases communicably coupled to the one or more servers, and one or more large language models communicably coupled to the one or more servers; automatically classifying one or more files or objects within the one or more databases based on one or more identifiers using the one or more servers; automatically summarizing the one or more files or objects using an autonomous AI agent on the one or more servers and communicably coupled to the one or more databases and the one or more large language models; automatically identifying people, organizations and issues within the one or more files or objects using the autonomous AI agent; and storing the summaries, and identified people, organizations and issues within the one or more databases.

The invention discloses a service monitoring data enhancement method, system and device based on LLM and retrieval enhancement generation and a storage medium, and relates to the technical field of service monitoring and exception handling. The method comprises the steps that user exception description is received, a webpage is retrieved through semantic expansion, and triple output structured data is extracted; and comparing with a local knowledge base, processing and storing the new content classification vector, and updating the index. Performing preliminary screening and reordering by combining exception description and an update library, taking a front list as a context to enable a large language model to generate analysis, evaluation and a strategy, and performing rule check and output; a result is verified, scene variants are generated according to exception types, after duplicate removal, the scene variants, strategies and metadata are structurally stored in a knowledge base, and continuous learning of a closed loop is completed; according to the method, high-quality and high-timeliness service exception training data and processing strategies can be automatically generated, the illusion problem of a large language model in actual operation and maintenance is effectively solved, and the fault response speed and the decision accuracy are improved.

This invention discloses an automatic update system for website article, tag, and category matching models, relating to the field of model update technology. It includes a data input module, a recommendation query unit, and a training unit. The data input module includes inputting query data and matching keyword data with data obtained from the database by the recommendation query unit to obtain matching data and corresponding articles. The database used is the Milvus database. This application optimizes the model and its performance, accelerating the training speed. Furthermore, this application uses the Milvus database, the Roformer model, and the prompt technique to solve the problem of requiring multiple manual training and restarts for a single site. Model updates are convenient and do not affect business operations. The same model can be adapted to data groups from multiple sites, and the problem of repeated training across multiple sites is solved.

The invention relates to the technical field of artificial intelligence, and provides a memory data processing and question answering method and device of a large language model, computer equipment and a readable storage medium. The memory data processing method of the large language model comprises the following steps: according to a memory summary template of a business field to which a to-be-processed dialogue record fragment of the large language model belongs, performing memory summary on the to-be-processed dialogue record fragment to obtain a to-be-processed memory stream; performing conflict detection on the to-be-processed memory stream and a memory stream in a memory database of the large language model; when a conflict detection result shows that the high-correlation memory flow of the to-be-processed memory flow exists in the memory database, determining a memory update content type of the to-be-processed memory flow; and according to the memory update content type, determining a corresponding processing measure so as to perform corresponding processing on the to-be-processed memory flow and the high-correlation memory flow. By adopting the method, the accuracy of model answering can be improved.

The invention relates to an intelligent collecting and processing method for open source intelligence data crawling, and belongs to the technical field of data processing. According to the method, after open-source information data is crawled by a traditional crawler method, an HTML tag exceeding the length of a context window of a large language model in the open-source information data is subjected to iterative segmentation from coarse to fine, and segmentation blocks with context relation are recombined into single segmentation blocks in each time of segmentation; and after each finally obtained segmentation block does not exceed the length of the context window, stopping segmentation iteration, and performing semantic recognition on a segmentation result by using a large language model to complete the collection of open source intelligence data. According to the method, the acquisition speed of the open source information data is ensured by a traditional crawler method, the recognition accuracy of the large language model on the HTML tag is ensured to the greatest extent on the basis of adaptive segmentation on the HTML tag, and finally, the acquisition speed and accuracy of the LLM-based crawler method on the open source information data can be improved.

The invention provides a double-server combined keyword dynamic searchable encryption method with forward and backward security, which is applied to a system comprising a data owner, a data user, a first server and a second server. Keywords and associated file identifiers are stored in each leaf node, information of non-leaf nodes is mapped to a non-distinguishable long filter, and through a random state updating mechanism, a keyword double-state array and a time factor trap door design, it is guaranteed that keyword joint search is achieved, and meanwhile, the information of the non-leaf nodes can be quickly and efficiently searched. And the forward safety and the backward safety are also ensured. According to the dual-server architecture, security index storage and trap door verification are separated, so that the problem of KPRP leakage is effectively relieved.

A system and method is presented that extracts snippets form web pages according to specially designed logic. The extracted snippets might be made relevant to, i.e., indexed by, a location and time / day applicability. Such snippets may be thought of as apps or services that are defined only when a mobile terminal is in a pre-defined geographical area at a certain time and day (e.g., as defined by a calendar of events). Extracted snippets are stored and made searchable. Methods and a system are described to control the display of snippet search results. Snippets may be selected by user or by programmed logic and executed on the mobile terminal or in remote servers without the need to download the app or source code associated with the snippet.

The application discloses a kind of Python model distributed online deployment method and system, data processing and feature engineering code converter designed by the application convert the logic of offline data processing and feature engineering of Python model training front-end into online Python code;Model engine designed by the application can quickly load various different kinds of Python models such as machine learning, deep learning, natural language processing etc.Through the Python model distributed deployment method and system provided by the application, Python model can be quickly deployed online, and can easily meet the requirements of high concurrency, high availability and fault tolerance etc.Algorithm engineers only need to spend time and effort on modeling, thereby improving the model development efficiency and shortening the model iteration cycle.

This application constructs a method, system, and storage medium for creating a digital twin of a power plant based on real-scene panoramic imagery. The method includes the following steps: Step S1: Acquire panoramic image data of the power plant site and perform panoramic image digitization processing to generate panoramic data; Step S2: Use a deep learning model to detect and identify equipment targets in the panoramic image data, extract equipment identification information, calculate the spatial location information of the equipment in the panoramic image data, and establish an association mapping based on the identification information and spatial location information. This application achieves low-cost, high-fidelity real-scene restoration through panoramic image digitization processing, achieves automatic equipment identification and annotation through a deep learning model, and achieves the fusion of visual data and ledger data through spatial location association mapping, thus realizing the rapid construction of a real-scene twin of a power plant.

This invention discloses a secure voice-based intelligent customer service system and interaction method based on a multi-agent architecture. The system includes an ASR (Automatic Speech Recognition) module, multi-layered security barriers, a main retrieval agent, a specialized retrieval agent cluster, a parallel retrieval and result aggregation module, an LLM (Limited Language Modeling) synthesis engine, and a TTS (Text-to-Speech) module. Through multi-agent system architecture and methodological design, this invention achieves modular processing, parallel retrieval tasks, and closed-loop security control. By enabling independent calls between input, output, agents, and inference stages via multi-layered security barriers, the system effectively mitigates risks such as prompt word injection, indirect injection attacks, harmful content generation, and information leakage. Sensitive information identification, access permission verification, and content compliance review are embedded in each stage of input, retrieval, information synthesis, and output to prevent sensitive data leakage and meet the application requirements of high-security business scenarios (such as finance, telecommunications, and government services).

The application provides a non-negative matrix factorization community detection method based on latent structure and attribute, and relates to the field of community network detection. The method mainly solves the problem that the existing method does not sufficiently utilize community network information. The method comprises the following steps: firstly, preprocessing community network data, and representing information in the network; calculating an adjacency matrix, an attribute matrix and a latent structure matrix; establishing a community detection model mainly based on direct topological structure and attribute information and supplemented by latent structure information; calculating an iterative updating rule; setting the number of iterations, initializing a community member matrix, two community-community matrices and the latent structure matrix, and adjusting a weight parameter; performing iteration to obtain a target matrix; and finally discovering a community according to the iteration result community member matrix.

The present invention relates to searching in digital information. In particular, the present invention relates to a software for searching in digital information, a system 200 comprising a plurality of such software, and a user device 100 connected to such system. An object of the present invention is to provide such a software, system 200 and user device 100 that can take advantage of existing horizontal search engines, vertical search engines and other data sources, to achieve a flexible and powerful search engine.

The invention provides a human-computer interaction method, electronic equipment and a storage medium, relates to the technical field of human-computer interaction, and is used for reducing user operation steps and improving user experience. The method comprises the following steps: in response to detection of a selection operation of a user on data on a current display screen, obtaining data content selected by the user, and receiving an operation instruction input by the user; calling the first large language model to perform processing operation indicated by the operation instruction on the data content to obtain a processing result of the data content; and displaying the processing result on the current display screen.

The invention discloses a distributed cloud storage optimization method oriented to large-scale data, which relates to the technical field of distributed cloud storage and comprises the following steps of: 1, generating a distribution decision, label mapping and time alignment number at an entrance according to a small object threshold and a late access score, and filing the distribution decision, the label mapping and the time alignment number; 2, storing a source object in a storage bucket, synthesizing a composite container according to the identifier, outputting a list file, an index segment and a pre-write log, and setting a source object retention window; 3, during reading, a cloud side firstly copies a source object to a heat collection barrel return link by a server side, and an object side performs container segmentation reading and triggers hot fragment extraction; 4, collecting a standard layer returning proportion or a hot layer returning proportion, delay and cost, performing closed-loop updating on parameters, and rewriting and recycling according to a combined triggering amount; according to the method, cross-cloud read-write collaboration, caliber consistency and traceability are realized, long-term cost is reduced, tail time delay is stabilized, and maintainability and expansibility are enhanced.

Methods and systems consistent with the disclosure include a method for generating a website, comprising: receiving, by a computer program executed on an electronic device, a request for a website; generating, by the computer program, a formatted request for the website to a microservice hub; identifying, by the microservice hub, one or more template parts based on the formatted request; determining, by the microservice hub, a site specific data to retrieve and sending a call to a site specific database for the site specific data; determining, by the microservice hub, a prompt to query a first machine learning engine based on an internal or external information content of the one or more template parts to fulfill a one or more prompts based on at least one template parts; and rendering, by the microservice hub, the website based on the prompts and the one or more template parts.

The invention provides a search method and device, equipment, a storage medium and a program product. The search method comprises the steps that multiple search results returned by multiple search engines for at least one query are obtained; the release time, the website domain name and the content feature of each search result are extracted; based on a pre-trained large language model, the release time, the website domain name and the content feature of each search result and the query corresponding to each search result, obtaining an evaluation result of each search engine; and determining a target search engine from the plurality of search engines based on the evaluation result of each search engine, and carrying out search based on the target search engine. By means of multi-dimensional features and a large language model, automatic evaluation of the search effect of the search engine is achieved, and evaluation objectivity, accuracy and comprehensiveness are improved; meanwhile, the engine is automatically selected for searching based on the evaluation result, and the hit rate of searching is increased.

The application provides a method and related device for communication between a native application and a webpage, and relates to the technical field of computers.In the application, when the webpage determines that a target function of the native application needs to be called, the function calling information is sent to the native application; the webpage receives the function callback result returned by the native application, the function callback result is determined and returned by the target configuration parameter defined by the corresponding business code of the native application under the corresponding operating system system, and the target configuration parameter defined by the business code under any operating system system is determined based on the system general communication protocol between the native application and the webpage.No matter what kind of operating system the native application runs under, the target configuration parameter defined by each business code is determined based on the system general communication protocol, that is, the configuration parameters defined in each business code are consistent, the workload of adaptation is reduced, and the adaptation cost is reduced.

The disclosure provides a click rate estimation model training method and device, and relates to the technical field of artificial intelligence. A specific embodiment of the method comprises: obtaining a preset sample set, the sample set comprising at least one sample; obtaining a click rate estimation network comprising a feature prior module and a model prior module, which is constructed in advance; performing the following training steps: selecting a sample from the sample set; inputting the sample into the click rate estimation network to obtain a click rate estimation value output by the click rate estimation network; calculating a feature loss value of a feature prior layer of the feature prior module and a model loss value common to a feature processing layer of the feature prior module and the model prior module; and in response to the click rate estimation network satisfying a training completion condition, taking the click rate estimation network as a click rate estimation model, which improves the generalization ability of the model.

The invention discloses a database query method, device and equipment based on a WebAPI interface and a storage medium, in the query method, WebAPI interface input parameters comprise an input parameter object and a task number, after a client sends a request message comprising the WebAPI interface input parameters, the WebAPI interface converts the input parameter object into dictionary data in a standard format, and the dictionary data in the standard format are stored in the database. The method comprises the following steps: acquiring a task number from a data center, acquiring a corresponding execution script from the data center according to the task number, assigning dictionary data to the acquired execution script, executing query of a database according to the execution script to obtain a result array, and finally converting the obtained result array into an output format and sending the output format to a client. The input parameters of the WebAPI interface are compatible with a plurality of interface requirements of different queries, different input parameters and different outputs by setting the input parameter objects and the task numbers, so that the design and development cost is effectively reduced while the personalized interface requirements are quickly responded, and code recompiling and redeployment online are not needed every time the requirements are changed.